Markov carpet embedded type feature selection method based on packaging

Abstract

The invention discloses a Markov carpet embedded type feature selection method based on packaging. The Markov carpet embedded type feature selection method is characterized by being carried out according to the following steps: 1, acquiring an optimal feature by using a five-fold cross validation method; 2, judging whether the optimal feature is a null set not; if so, finishing feature selection; otherwise, updating a feature subset; 3, deleting redundancy features by using a Markov carpet method so as to update a feature vector; and 4, judging whether the feature vector is a null set not; if so, finishing the feature selection; otherwise, repeating the step 2. By virtue of adopting the Markov carpet embedded type feature selection method, the high-quality feature subset can be obtained; and meanwhile, the time complexity of the feature selection method based on the packaging is reduced so that relatively good classification performance and time performance can be obtained.

Description

technical field [0001] The invention belongs to the field of data mining, in particular to a Markov blanket-embedded feature selection method based on encapsulation. Background technique [0002] Feature selection, as a data preprocessing technique, is widely used in machine learning and data mining tasks, such as classification, regression, and clustering. When the original feature space of the data includes irrelevant or redundant features with the target task, classifiers built on the entire feature space often have poor performance, for example, Naive Bayesian classifiers are sensitive to redundant features. The purpose of feature selection is to apply an effective feature selection method to select a set of discriminative features from the original feature space. An effective feature selection method can not only reduce the dimensionality of the original feature space, but also reduce the training time of the classifier and improve its generalization ability. More impo...

AI Technical Summary

Problems solved by technology

[0006] (2) Redundant features in the candidate feature set cannot be quickly identified, and these redundant features are kept in the candidate feature set until the end of the feature selection method, resulting in repeated evaluation of these redundant features

Images